forked from suyu/suyu
Merge pull request #9453 from ameerj/scratch-vector
common: Add ScratchBuffer Class
This commit is contained in:
commit
3e6850f00b
14 changed files with 370 additions and 56 deletions
|
@ -78,6 +78,7 @@ add_library(common STATIC
|
||||||
logging/types.h
|
logging/types.h
|
||||||
lz4_compression.cpp
|
lz4_compression.cpp
|
||||||
lz4_compression.h
|
lz4_compression.h
|
||||||
|
make_unique_for_overwrite.h
|
||||||
math_util.h
|
math_util.h
|
||||||
memory_detect.cpp
|
memory_detect.cpp
|
||||||
memory_detect.h
|
memory_detect.h
|
||||||
|
@ -101,6 +102,7 @@ add_library(common STATIC
|
||||||
${CMAKE_CURRENT_BINARY_DIR}/scm_rev.cpp
|
${CMAKE_CURRENT_BINARY_DIR}/scm_rev.cpp
|
||||||
scm_rev.h
|
scm_rev.h
|
||||||
scope_exit.h
|
scope_exit.h
|
||||||
|
scratch_buffer.h
|
||||||
settings.cpp
|
settings.cpp
|
||||||
settings.h
|
settings.h
|
||||||
settings_input.cpp
|
settings_input.cpp
|
||||||
|
|
25
src/common/make_unique_for_overwrite.h
Normal file
25
src/common/make_unique_for_overwrite.h
Normal file
|
@ -0,0 +1,25 @@
|
||||||
|
// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
|
||||||
|
// SPDX-License-Identifier: GPL-2.0-or-later
|
||||||
|
|
||||||
|
#pragma once
|
||||||
|
|
||||||
|
#include <memory>
|
||||||
|
#include <type_traits>
|
||||||
|
|
||||||
|
namespace Common {
|
||||||
|
|
||||||
|
template <class T>
|
||||||
|
requires(!std::is_array_v<T>) std::unique_ptr<T> make_unique_for_overwrite() {
|
||||||
|
return std::unique_ptr<T>(new T);
|
||||||
|
}
|
||||||
|
|
||||||
|
template <class T>
|
||||||
|
requires std::is_unbounded_array_v<T> std::unique_ptr<T> make_unique_for_overwrite(std::size_t n) {
|
||||||
|
return std::unique_ptr<T>(new std::remove_extent_t<T>[n]);
|
||||||
|
}
|
||||||
|
|
||||||
|
template <class T, class... Args>
|
||||||
|
requires std::is_bounded_array_v<T>
|
||||||
|
void make_unique_for_overwrite(Args&&...) = delete;
|
||||||
|
|
||||||
|
} // namespace Common
|
95
src/common/scratch_buffer.h
Normal file
95
src/common/scratch_buffer.h
Normal file
|
@ -0,0 +1,95 @@
|
||||||
|
// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
|
||||||
|
// SPDX-License-Identifier: GPL-2.0-or-later
|
||||||
|
|
||||||
|
#pragma once
|
||||||
|
|
||||||
|
#include "common/make_unique_for_overwrite.h"
|
||||||
|
|
||||||
|
namespace Common {
|
||||||
|
|
||||||
|
/**
|
||||||
|
* ScratchBuffer class
|
||||||
|
* This class creates a default initialized heap allocated buffer for cases such as intermediate
|
||||||
|
* buffers being copied into entirely, where value initializing members during allocation or resize
|
||||||
|
* is redundant.
|
||||||
|
*/
|
||||||
|
template <typename T>
|
||||||
|
class ScratchBuffer {
|
||||||
|
public:
|
||||||
|
ScratchBuffer() = default;
|
||||||
|
|
||||||
|
explicit ScratchBuffer(size_t initial_capacity)
|
||||||
|
: last_requested_size{initial_capacity}, buffer_capacity{initial_capacity},
|
||||||
|
buffer{Common::make_unique_for_overwrite<T[]>(initial_capacity)} {}
|
||||||
|
|
||||||
|
~ScratchBuffer() = default;
|
||||||
|
|
||||||
|
/// This will only grow the buffer's capacity if size is greater than the current capacity.
|
||||||
|
/// The previously held data will remain intact.
|
||||||
|
void resize(size_t size) {
|
||||||
|
if (size > buffer_capacity) {
|
||||||
|
auto new_buffer = Common::make_unique_for_overwrite<T[]>(size);
|
||||||
|
std::move(buffer.get(), buffer.get() + buffer_capacity, new_buffer.get());
|
||||||
|
buffer = std::move(new_buffer);
|
||||||
|
buffer_capacity = size;
|
||||||
|
}
|
||||||
|
last_requested_size = size;
|
||||||
|
}
|
||||||
|
|
||||||
|
/// This will only grow the buffer's capacity if size is greater than the current capacity.
|
||||||
|
/// The previously held data will be destroyed if a reallocation occurs.
|
||||||
|
void resize_destructive(size_t size) {
|
||||||
|
if (size > buffer_capacity) {
|
||||||
|
buffer_capacity = size;
|
||||||
|
buffer = Common::make_unique_for_overwrite<T[]>(buffer_capacity);
|
||||||
|
}
|
||||||
|
last_requested_size = size;
|
||||||
|
}
|
||||||
|
|
||||||
|
[[nodiscard]] T* data() noexcept {
|
||||||
|
return buffer.get();
|
||||||
|
}
|
||||||
|
|
||||||
|
[[nodiscard]] const T* data() const noexcept {
|
||||||
|
return buffer.get();
|
||||||
|
}
|
||||||
|
|
||||||
|
[[nodiscard]] T* begin() noexcept {
|
||||||
|
return data();
|
||||||
|
}
|
||||||
|
|
||||||
|
[[nodiscard]] const T* begin() const noexcept {
|
||||||
|
return data();
|
||||||
|
}
|
||||||
|
|
||||||
|
[[nodiscard]] T* end() noexcept {
|
||||||
|
return data() + last_requested_size;
|
||||||
|
}
|
||||||
|
|
||||||
|
[[nodiscard]] const T* end() const noexcept {
|
||||||
|
return data() + last_requested_size;
|
||||||
|
}
|
||||||
|
|
||||||
|
[[nodiscard]] T& operator[](size_t i) {
|
||||||
|
return buffer[i];
|
||||||
|
}
|
||||||
|
|
||||||
|
[[nodiscard]] const T& operator[](size_t i) const {
|
||||||
|
return buffer[i];
|
||||||
|
}
|
||||||
|
|
||||||
|
[[nodiscard]] size_t size() const noexcept {
|
||||||
|
return last_requested_size;
|
||||||
|
}
|
||||||
|
|
||||||
|
[[nodiscard]] size_t capacity() const noexcept {
|
||||||
|
return buffer_capacity;
|
||||||
|
}
|
||||||
|
|
||||||
|
private:
|
||||||
|
size_t last_requested_size{};
|
||||||
|
size_t buffer_capacity{};
|
||||||
|
std::unique_ptr<T[]> buffer{};
|
||||||
|
};
|
||||||
|
|
||||||
|
} // namespace Common
|
|
@ -8,6 +8,7 @@ add_executable(tests
|
||||||
common/host_memory.cpp
|
common/host_memory.cpp
|
||||||
common/param_package.cpp
|
common/param_package.cpp
|
||||||
common/ring_buffer.cpp
|
common/ring_buffer.cpp
|
||||||
|
common/scratch_buffer.cpp
|
||||||
common/unique_function.cpp
|
common/unique_function.cpp
|
||||||
core/core_timing.cpp
|
core/core_timing.cpp
|
||||||
core/internal_network/network.cpp
|
core/internal_network/network.cpp
|
||||||
|
|
199
src/tests/common/scratch_buffer.cpp
Normal file
199
src/tests/common/scratch_buffer.cpp
Normal file
|
@ -0,0 +1,199 @@
|
||||||
|
// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
|
||||||
|
// SPDX-License-Identifier: GPL-2.0-or-later
|
||||||
|
|
||||||
|
#include <algorithm>
|
||||||
|
#include <array>
|
||||||
|
#include <span>
|
||||||
|
#include <catch2/catch.hpp>
|
||||||
|
#include "common/common_types.h"
|
||||||
|
#include "common/scratch_buffer.h"
|
||||||
|
|
||||||
|
namespace Common {
|
||||||
|
|
||||||
|
TEST_CASE("ScratchBuffer: Basic Test", "[common]") {
|
||||||
|
ScratchBuffer<u8> buf;
|
||||||
|
|
||||||
|
REQUIRE(buf.size() == 0U);
|
||||||
|
REQUIRE(buf.capacity() == 0U);
|
||||||
|
|
||||||
|
std::array<u8, 10> payload;
|
||||||
|
payload.fill(66);
|
||||||
|
|
||||||
|
buf.resize(payload.size());
|
||||||
|
REQUIRE(buf.size() == payload.size());
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
std::memcpy(buf.data(), payload.data(), payload.size());
|
||||||
|
for (size_t i = 0; i < payload.size(); ++i) {
|
||||||
|
REQUIRE(buf[i] == payload[i]);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
TEST_CASE("ScratchBuffer: resize_destructive Grow", "[common]") {
|
||||||
|
std::array<u8, 10> payload;
|
||||||
|
payload.fill(66);
|
||||||
|
|
||||||
|
ScratchBuffer<u8> buf(payload.size());
|
||||||
|
REQUIRE(buf.size() == payload.size());
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
// Increasing the size should reallocate the buffer
|
||||||
|
buf.resize_destructive(payload.size() * 2);
|
||||||
|
REQUIRE(buf.size() == payload.size() * 2);
|
||||||
|
REQUIRE(buf.capacity() == payload.size() * 2);
|
||||||
|
|
||||||
|
// Since the buffer is not value initialized, reading its data will be garbage
|
||||||
|
}
|
||||||
|
|
||||||
|
TEST_CASE("ScratchBuffer: resize_destructive Shrink", "[common]") {
|
||||||
|
std::array<u8, 10> payload;
|
||||||
|
payload.fill(66);
|
||||||
|
|
||||||
|
ScratchBuffer<u8> buf(payload.size());
|
||||||
|
REQUIRE(buf.size() == payload.size());
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
std::memcpy(buf.data(), payload.data(), payload.size());
|
||||||
|
for (size_t i = 0; i < payload.size(); ++i) {
|
||||||
|
REQUIRE(buf[i] == payload[i]);
|
||||||
|
}
|
||||||
|
|
||||||
|
// Decreasing the size should not cause a buffer reallocation
|
||||||
|
// This can be tested by ensuring the buffer capacity and data has not changed,
|
||||||
|
buf.resize_destructive(1U);
|
||||||
|
REQUIRE(buf.size() == 1U);
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
for (size_t i = 0; i < payload.size(); ++i) {
|
||||||
|
REQUIRE(buf[i] == payload[i]);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
TEST_CASE("ScratchBuffer: resize Grow u8", "[common]") {
|
||||||
|
std::array<u8, 10> payload;
|
||||||
|
payload.fill(66);
|
||||||
|
|
||||||
|
ScratchBuffer<u8> buf(payload.size());
|
||||||
|
REQUIRE(buf.size() == payload.size());
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
std::memcpy(buf.data(), payload.data(), payload.size());
|
||||||
|
for (size_t i = 0; i < payload.size(); ++i) {
|
||||||
|
REQUIRE(buf[i] == payload[i]);
|
||||||
|
}
|
||||||
|
|
||||||
|
// Increasing the size should reallocate the buffer
|
||||||
|
buf.resize(payload.size() * 2);
|
||||||
|
REQUIRE(buf.size() == payload.size() * 2);
|
||||||
|
REQUIRE(buf.capacity() == payload.size() * 2);
|
||||||
|
|
||||||
|
// resize() keeps the previous data intact
|
||||||
|
for (size_t i = 0; i < payload.size(); ++i) {
|
||||||
|
REQUIRE(buf[i] == payload[i]);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
TEST_CASE("ScratchBuffer: resize Grow u64", "[common]") {
|
||||||
|
std::array<u64, 10> payload;
|
||||||
|
payload.fill(6666);
|
||||||
|
|
||||||
|
ScratchBuffer<u64> buf(payload.size());
|
||||||
|
REQUIRE(buf.size() == payload.size());
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
std::memcpy(buf.data(), payload.data(), payload.size() * sizeof(u64));
|
||||||
|
for (size_t i = 0; i < payload.size(); ++i) {
|
||||||
|
REQUIRE(buf[i] == payload[i]);
|
||||||
|
}
|
||||||
|
|
||||||
|
// Increasing the size should reallocate the buffer
|
||||||
|
buf.resize(payload.size() * 2);
|
||||||
|
REQUIRE(buf.size() == payload.size() * 2);
|
||||||
|
REQUIRE(buf.capacity() == payload.size() * 2);
|
||||||
|
|
||||||
|
// resize() keeps the previous data intact
|
||||||
|
for (size_t i = 0; i < payload.size(); ++i) {
|
||||||
|
REQUIRE(buf[i] == payload[i]);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
TEST_CASE("ScratchBuffer: resize Shrink", "[common]") {
|
||||||
|
std::array<u8, 10> payload;
|
||||||
|
payload.fill(66);
|
||||||
|
|
||||||
|
ScratchBuffer<u8> buf(payload.size());
|
||||||
|
REQUIRE(buf.size() == payload.size());
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
std::memcpy(buf.data(), payload.data(), payload.size());
|
||||||
|
for (size_t i = 0; i < payload.size(); ++i) {
|
||||||
|
REQUIRE(buf[i] == payload[i]);
|
||||||
|
}
|
||||||
|
|
||||||
|
// Decreasing the size should not cause a buffer reallocation
|
||||||
|
// This can be tested by ensuring the buffer capacity and data has not changed,
|
||||||
|
buf.resize(1U);
|
||||||
|
REQUIRE(buf.size() == 1U);
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
for (size_t i = 0; i < payload.size(); ++i) {
|
||||||
|
REQUIRE(buf[i] == payload[i]);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
TEST_CASE("ScratchBuffer: Span Size", "[common]") {
|
||||||
|
std::array<u8, 10> payload;
|
||||||
|
payload.fill(66);
|
||||||
|
|
||||||
|
ScratchBuffer<u8> buf(payload.size());
|
||||||
|
REQUIRE(buf.size() == payload.size());
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
std::memcpy(buf.data(), payload.data(), payload.size());
|
||||||
|
for (size_t i = 0; i < payload.size(); ++i) {
|
||||||
|
REQUIRE(buf[i] == payload[i]);
|
||||||
|
}
|
||||||
|
|
||||||
|
buf.resize(3U);
|
||||||
|
REQUIRE(buf.size() == 3U);
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
const auto buf_span = std::span<u8>(buf);
|
||||||
|
// The span size is the last requested size of the buffer, not its capacity
|
||||||
|
REQUIRE(buf_span.size() == buf.size());
|
||||||
|
|
||||||
|
for (size_t i = 0; i < buf_span.size(); ++i) {
|
||||||
|
REQUIRE(buf_span[i] == buf[i]);
|
||||||
|
REQUIRE(buf_span[i] == payload[i]);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
TEST_CASE("ScratchBuffer: Span Writes", "[common]") {
|
||||||
|
std::array<u8, 10> payload;
|
||||||
|
payload.fill(66);
|
||||||
|
|
||||||
|
ScratchBuffer<u8> buf(payload.size());
|
||||||
|
REQUIRE(buf.size() == payload.size());
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
std::memcpy(buf.data(), payload.data(), payload.size());
|
||||||
|
for (size_t i = 0; i < payload.size(); ++i) {
|
||||||
|
REQUIRE(buf[i] == payload[i]);
|
||||||
|
}
|
||||||
|
|
||||||
|
buf.resize(3U);
|
||||||
|
REQUIRE(buf.size() == 3U);
|
||||||
|
REQUIRE(buf.capacity() == payload.size());
|
||||||
|
|
||||||
|
const auto buf_span = std::span<u8>(buf);
|
||||||
|
REQUIRE(buf_span.size() == buf.size());
|
||||||
|
|
||||||
|
for (size_t i = 0; i < buf_span.size(); ++i) {
|
||||||
|
const auto new_value = static_cast<u8>(i + 1U);
|
||||||
|
// Writes to a span of the scratch buffer will propogate to the buffer itself
|
||||||
|
buf_span[i] = new_value;
|
||||||
|
REQUIRE(buf[i] == new_value);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
} // namespace Common
|
|
@ -20,6 +20,7 @@
|
||||||
#include "common/lru_cache.h"
|
#include "common/lru_cache.h"
|
||||||
#include "common/microprofile.h"
|
#include "common/microprofile.h"
|
||||||
#include "common/polyfill_ranges.h"
|
#include "common/polyfill_ranges.h"
|
||||||
|
#include "common/scratch_buffer.h"
|
||||||
#include "common/settings.h"
|
#include "common/settings.h"
|
||||||
#include "core/memory.h"
|
#include "core/memory.h"
|
||||||
#include "video_core/buffer_cache/buffer_base.h"
|
#include "video_core/buffer_cache/buffer_base.h"
|
||||||
|
@ -422,8 +423,7 @@ private:
|
||||||
IntervalSet common_ranges;
|
IntervalSet common_ranges;
|
||||||
std::deque<IntervalSet> committed_ranges;
|
std::deque<IntervalSet> committed_ranges;
|
||||||
|
|
||||||
size_t immediate_buffer_capacity = 0;
|
Common::ScratchBuffer<u8> immediate_buffer_alloc;
|
||||||
std::unique_ptr<u8[]> immediate_buffer_alloc;
|
|
||||||
|
|
||||||
struct LRUItemParams {
|
struct LRUItemParams {
|
||||||
using ObjectType = BufferId;
|
using ObjectType = BufferId;
|
||||||
|
@ -1926,11 +1926,8 @@ std::span<const u8> BufferCache<P>::ImmediateBufferWithData(VAddr cpu_addr, size
|
||||||
|
|
||||||
template <class P>
|
template <class P>
|
||||||
std::span<u8> BufferCache<P>::ImmediateBuffer(size_t wanted_capacity) {
|
std::span<u8> BufferCache<P>::ImmediateBuffer(size_t wanted_capacity) {
|
||||||
if (wanted_capacity > immediate_buffer_capacity) {
|
immediate_buffer_alloc.resize_destructive(wanted_capacity);
|
||||||
immediate_buffer_capacity = wanted_capacity;
|
return std::span<u8>(immediate_buffer_alloc.data(), wanted_capacity);
|
||||||
immediate_buffer_alloc = std::make_unique<u8[]>(wanted_capacity);
|
|
||||||
}
|
|
||||||
return std::span<u8>(immediate_buffer_alloc.get(), wanted_capacity);
|
|
||||||
}
|
}
|
||||||
|
|
||||||
template <class P>
|
template <class P>
|
||||||
|
|
|
@ -56,7 +56,7 @@ bool DmaPusher::Step() {
|
||||||
|
|
||||||
if (command_list.prefetch_command_list.size()) {
|
if (command_list.prefetch_command_list.size()) {
|
||||||
// Prefetched command list from nvdrv, used for things like synchronization
|
// Prefetched command list from nvdrv, used for things like synchronization
|
||||||
command_headers = std::move(command_list.prefetch_command_list);
|
ProcessCommands(command_list.prefetch_command_list);
|
||||||
dma_pushbuffer.pop();
|
dma_pushbuffer.pop();
|
||||||
} else {
|
} else {
|
||||||
const CommandListHeader command_list_header{
|
const CommandListHeader command_list_header{
|
||||||
|
@ -74,7 +74,7 @@ bool DmaPusher::Step() {
|
||||||
}
|
}
|
||||||
|
|
||||||
// Push buffer non-empty, read a word
|
// Push buffer non-empty, read a word
|
||||||
command_headers.resize(command_list_header.size);
|
command_headers.resize_destructive(command_list_header.size);
|
||||||
if (Settings::IsGPULevelHigh()) {
|
if (Settings::IsGPULevelHigh()) {
|
||||||
memory_manager.ReadBlock(dma_get, command_headers.data(),
|
memory_manager.ReadBlock(dma_get, command_headers.data(),
|
||||||
command_list_header.size * sizeof(u32));
|
command_list_header.size * sizeof(u32));
|
||||||
|
@ -82,16 +82,21 @@ bool DmaPusher::Step() {
|
||||||
memory_manager.ReadBlockUnsafe(dma_get, command_headers.data(),
|
memory_manager.ReadBlockUnsafe(dma_get, command_headers.data(),
|
||||||
command_list_header.size * sizeof(u32));
|
command_list_header.size * sizeof(u32));
|
||||||
}
|
}
|
||||||
|
ProcessCommands(command_headers);
|
||||||
}
|
}
|
||||||
for (std::size_t index = 0; index < command_headers.size();) {
|
|
||||||
const CommandHeader& command_header = command_headers[index];
|
return true;
|
||||||
|
}
|
||||||
|
|
||||||
|
void DmaPusher::ProcessCommands(std::span<const CommandHeader> commands) {
|
||||||
|
for (std::size_t index = 0; index < commands.size();) {
|
||||||
|
const CommandHeader& command_header = commands[index];
|
||||||
|
|
||||||
if (dma_state.method_count) {
|
if (dma_state.method_count) {
|
||||||
// Data word of methods command
|
// Data word of methods command
|
||||||
if (dma_state.non_incrementing) {
|
if (dma_state.non_incrementing) {
|
||||||
const u32 max_write = static_cast<u32>(
|
const u32 max_write = static_cast<u32>(
|
||||||
std::min<std::size_t>(index + dma_state.method_count, command_headers.size()) -
|
std::min<std::size_t>(index + dma_state.method_count, commands.size()) - index);
|
||||||
index);
|
|
||||||
CallMultiMethod(&command_header.argument, max_write);
|
CallMultiMethod(&command_header.argument, max_write);
|
||||||
dma_state.method_count -= max_write;
|
dma_state.method_count -= max_write;
|
||||||
dma_state.is_last_call = true;
|
dma_state.is_last_call = true;
|
||||||
|
@ -142,8 +147,6 @@ bool DmaPusher::Step() {
|
||||||
}
|
}
|
||||||
index++;
|
index++;
|
||||||
}
|
}
|
||||||
|
|
||||||
return true;
|
|
||||||
}
|
}
|
||||||
|
|
||||||
void DmaPusher::SetState(const CommandHeader& command_header) {
|
void DmaPusher::SetState(const CommandHeader& command_header) {
|
||||||
|
|
|
@ -4,11 +4,13 @@
|
||||||
#pragma once
|
#pragma once
|
||||||
|
|
||||||
#include <array>
|
#include <array>
|
||||||
|
#include <span>
|
||||||
#include <vector>
|
#include <vector>
|
||||||
#include <queue>
|
#include <queue>
|
||||||
|
|
||||||
#include "common/bit_field.h"
|
#include "common/bit_field.h"
|
||||||
#include "common/common_types.h"
|
#include "common/common_types.h"
|
||||||
|
#include "common/scratch_buffer.h"
|
||||||
#include "video_core/engines/engine_interface.h"
|
#include "video_core/engines/engine_interface.h"
|
||||||
#include "video_core/engines/puller.h"
|
#include "video_core/engines/puller.h"
|
||||||
|
|
||||||
|
@ -136,13 +138,15 @@ private:
|
||||||
static constexpr u32 non_puller_methods = 0x40;
|
static constexpr u32 non_puller_methods = 0x40;
|
||||||
static constexpr u32 max_subchannels = 8;
|
static constexpr u32 max_subchannels = 8;
|
||||||
bool Step();
|
bool Step();
|
||||||
|
void ProcessCommands(std::span<const CommandHeader> commands);
|
||||||
|
|
||||||
void SetState(const CommandHeader& command_header);
|
void SetState(const CommandHeader& command_header);
|
||||||
|
|
||||||
void CallMethod(u32 argument) const;
|
void CallMethod(u32 argument) const;
|
||||||
void CallMultiMethod(const u32* base_start, u32 num_methods) const;
|
void CallMultiMethod(const u32* base_start, u32 num_methods) const;
|
||||||
|
|
||||||
std::vector<CommandHeader> command_headers; ///< Buffer for list of commands fetched at once
|
Common::ScratchBuffer<CommandHeader>
|
||||||
|
command_headers; ///< Buffer for list of commands fetched at once
|
||||||
|
|
||||||
std::queue<CommandList> dma_pushbuffer; ///< Queue of command lists to be processed
|
std::queue<CommandList> dma_pushbuffer; ///< Queue of command lists to be processed
|
||||||
std::size_t dma_pushbuffer_subindex{}; ///< Index within a command list within the pushbuffer
|
std::size_t dma_pushbuffer_subindex{}; ///< Index within a command list within the pushbuffer
|
||||||
|
@ -159,7 +163,7 @@ private:
|
||||||
DmaState dma_state{};
|
DmaState dma_state{};
|
||||||
bool dma_increment_once{};
|
bool dma_increment_once{};
|
||||||
|
|
||||||
bool ib_enable{true}; ///< IB mode enabled
|
const bool ib_enable{true}; ///< IB mode enabled
|
||||||
|
|
||||||
std::array<Engines::EngineInterface*, max_subchannels> subchannels{};
|
std::array<Engines::EngineInterface*, max_subchannels> subchannels{};
|
||||||
|
|
||||||
|
|
|
@ -24,7 +24,7 @@ void State::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
|
||||||
void State::ProcessExec(const bool is_linear_) {
|
void State::ProcessExec(const bool is_linear_) {
|
||||||
write_offset = 0;
|
write_offset = 0;
|
||||||
copy_size = regs.line_length_in * regs.line_count;
|
copy_size = regs.line_length_in * regs.line_count;
|
||||||
inner_buffer.resize(copy_size);
|
inner_buffer.resize_destructive(copy_size);
|
||||||
is_linear = is_linear_;
|
is_linear = is_linear_;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
@ -70,7 +70,7 @@ void State::ProcessData(std::span<const u8> read_buffer) {
|
||||||
const std::size_t dst_size = Tegra::Texture::CalculateSize(
|
const std::size_t dst_size = Tegra::Texture::CalculateSize(
|
||||||
true, bytes_per_pixel, width, regs.dest.height, regs.dest.depth,
|
true, bytes_per_pixel, width, regs.dest.height, regs.dest.depth,
|
||||||
regs.dest.BlockHeight(), regs.dest.BlockDepth());
|
regs.dest.BlockHeight(), regs.dest.BlockDepth());
|
||||||
tmp_buffer.resize(dst_size);
|
tmp_buffer.resize_destructive(dst_size);
|
||||||
memory_manager.ReadBlock(address, tmp_buffer.data(), dst_size);
|
memory_manager.ReadBlock(address, tmp_buffer.data(), dst_size);
|
||||||
Tegra::Texture::SwizzleSubrect(tmp_buffer, read_buffer, bytes_per_pixel, width,
|
Tegra::Texture::SwizzleSubrect(tmp_buffer, read_buffer, bytes_per_pixel, width,
|
||||||
regs.dest.height, regs.dest.depth, x_offset, regs.dest.y,
|
regs.dest.height, regs.dest.depth, x_offset, regs.dest.y,
|
||||||
|
|
|
@ -4,9 +4,10 @@
|
||||||
#pragma once
|
#pragma once
|
||||||
|
|
||||||
#include <span>
|
#include <span>
|
||||||
#include <vector>
|
|
||||||
#include "common/bit_field.h"
|
#include "common/bit_field.h"
|
||||||
#include "common/common_types.h"
|
#include "common/common_types.h"
|
||||||
|
#include "common/scratch_buffer.h"
|
||||||
|
|
||||||
namespace Tegra {
|
namespace Tegra {
|
||||||
class MemoryManager;
|
class MemoryManager;
|
||||||
|
@ -73,8 +74,8 @@ private:
|
||||||
|
|
||||||
u32 write_offset = 0;
|
u32 write_offset = 0;
|
||||||
u32 copy_size = 0;
|
u32 copy_size = 0;
|
||||||
std::vector<u8> inner_buffer;
|
Common::ScratchBuffer<u8> inner_buffer;
|
||||||
std::vector<u8> tmp_buffer;
|
Common::ScratchBuffer<u8> tmp_buffer;
|
||||||
bool is_linear = false;
|
bool is_linear = false;
|
||||||
Registers& regs;
|
Registers& regs;
|
||||||
MemoryManager& memory_manager;
|
MemoryManager& memory_manager;
|
||||||
|
|
|
@ -184,12 +184,8 @@ void MaxwellDMA::CopyBlockLinearToPitch() {
|
||||||
const size_t src_size =
|
const size_t src_size =
|
||||||
CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth);
|
CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth);
|
||||||
|
|
||||||
if (read_buffer.size() < src_size) {
|
read_buffer.resize_destructive(src_size);
|
||||||
read_buffer.resize(src_size);
|
write_buffer.resize_destructive(dst_size);
|
||||||
}
|
|
||||||
if (write_buffer.size() < dst_size) {
|
|
||||||
write_buffer.resize(dst_size);
|
|
||||||
}
|
|
||||||
|
|
||||||
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
|
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
|
||||||
memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size);
|
memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size);
|
||||||
|
@ -235,12 +231,8 @@ void MaxwellDMA::CopyPitchToBlockLinear() {
|
||||||
CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth);
|
CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth);
|
||||||
const size_t src_size = static_cast<size_t>(regs.pitch_in) * regs.line_count;
|
const size_t src_size = static_cast<size_t>(regs.pitch_in) * regs.line_count;
|
||||||
|
|
||||||
if (read_buffer.size() < src_size) {
|
read_buffer.resize_destructive(src_size);
|
||||||
read_buffer.resize(src_size);
|
write_buffer.resize_destructive(dst_size);
|
||||||
}
|
|
||||||
if (write_buffer.size() < dst_size) {
|
|
||||||
write_buffer.resize(dst_size);
|
|
||||||
}
|
|
||||||
|
|
||||||
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
|
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
|
||||||
if (Settings::IsGPULevelExtreme()) {
|
if (Settings::IsGPULevelExtreme()) {
|
||||||
|
@ -269,12 +261,8 @@ void MaxwellDMA::FastCopyBlockLinearToPitch() {
|
||||||
pos_x = pos_x % x_in_gob;
|
pos_x = pos_x % x_in_gob;
|
||||||
pos_y = pos_y % 8;
|
pos_y = pos_y % 8;
|
||||||
|
|
||||||
if (read_buffer.size() < src_size) {
|
read_buffer.resize_destructive(src_size);
|
||||||
read_buffer.resize(src_size);
|
write_buffer.resize_destructive(dst_size);
|
||||||
}
|
|
||||||
if (write_buffer.size() < dst_size) {
|
|
||||||
write_buffer.resize(dst_size);
|
|
||||||
}
|
|
||||||
|
|
||||||
if (Settings::IsGPULevelExtreme()) {
|
if (Settings::IsGPULevelExtreme()) {
|
||||||
memory_manager.ReadBlock(regs.offset_in + offset, read_buffer.data(), src_size);
|
memory_manager.ReadBlock(regs.offset_in + offset, read_buffer.data(), src_size);
|
||||||
|
@ -333,14 +321,10 @@ void MaxwellDMA::CopyBlockLinearToBlockLinear() {
|
||||||
const u32 pitch = x_elements * bytes_per_pixel;
|
const u32 pitch = x_elements * bytes_per_pixel;
|
||||||
const size_t mid_buffer_size = pitch * regs.line_count;
|
const size_t mid_buffer_size = pitch * regs.line_count;
|
||||||
|
|
||||||
if (read_buffer.size() < src_size) {
|
read_buffer.resize_destructive(src_size);
|
||||||
read_buffer.resize(src_size);
|
write_buffer.resize_destructive(dst_size);
|
||||||
}
|
|
||||||
if (write_buffer.size() < dst_size) {
|
|
||||||
write_buffer.resize(dst_size);
|
|
||||||
}
|
|
||||||
|
|
||||||
intermediate_buffer.resize(mid_buffer_size);
|
intermediate_buffer.resize_destructive(mid_buffer_size);
|
||||||
|
|
||||||
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
|
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
|
||||||
memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size);
|
memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size);
|
||||||
|
|
|
@ -6,8 +6,10 @@
|
||||||
#include <array>
|
#include <array>
|
||||||
#include <cstddef>
|
#include <cstddef>
|
||||||
#include <vector>
|
#include <vector>
|
||||||
|
|
||||||
#include "common/bit_field.h"
|
#include "common/bit_field.h"
|
||||||
#include "common/common_types.h"
|
#include "common/common_types.h"
|
||||||
|
#include "common/scratch_buffer.h"
|
||||||
#include "video_core/engines/engine_interface.h"
|
#include "video_core/engines/engine_interface.h"
|
||||||
|
|
||||||
namespace Core {
|
namespace Core {
|
||||||
|
@ -234,9 +236,9 @@ private:
|
||||||
MemoryManager& memory_manager;
|
MemoryManager& memory_manager;
|
||||||
VideoCore::RasterizerInterface* rasterizer = nullptr;
|
VideoCore::RasterizerInterface* rasterizer = nullptr;
|
||||||
|
|
||||||
std::vector<u8> read_buffer;
|
Common::ScratchBuffer<u8> read_buffer;
|
||||||
std::vector<u8> write_buffer;
|
Common::ScratchBuffer<u8> write_buffer;
|
||||||
std::vector<u8> intermediate_buffer;
|
Common::ScratchBuffer<u8> intermediate_buffer;
|
||||||
|
|
||||||
static constexpr std::size_t NUM_REGS = 0x800;
|
static constexpr std::size_t NUM_REGS = 0x800;
|
||||||
struct Regs {
|
struct Regs {
|
||||||
|
|
|
@ -155,7 +155,7 @@ void Vic::WriteRGBFrame(const AVFrame* frame, const VicConfig& config) {
|
||||||
// swizzle pitch linear to block linear
|
// swizzle pitch linear to block linear
|
||||||
const u32 block_height = static_cast<u32>(config.block_linear_height_log2);
|
const u32 block_height = static_cast<u32>(config.block_linear_height_log2);
|
||||||
const auto size = Texture::CalculateSize(true, 4, width, height, 1, block_height, 0);
|
const auto size = Texture::CalculateSize(true, 4, width, height, 1, block_height, 0);
|
||||||
luma_buffer.resize(size);
|
luma_buffer.resize_destructive(size);
|
||||||
std::span<const u8> frame_buff(converted_frame_buf_addr, 4 * width * height);
|
std::span<const u8> frame_buff(converted_frame_buf_addr, 4 * width * height);
|
||||||
Texture::SwizzleSubrect(luma_buffer, frame_buff, 4, width, height, 1, 0, 0, width, height,
|
Texture::SwizzleSubrect(luma_buffer, frame_buff, 4, width, height, 1, 0, 0, width, height,
|
||||||
block_height, 0, width * 4);
|
block_height, 0, width * 4);
|
||||||
|
@ -181,8 +181,8 @@ void Vic::WriteYUVFrame(const AVFrame* frame, const VicConfig& config) {
|
||||||
|
|
||||||
const auto stride = static_cast<size_t>(frame->linesize[0]);
|
const auto stride = static_cast<size_t>(frame->linesize[0]);
|
||||||
|
|
||||||
luma_buffer.resize(aligned_width * surface_height);
|
luma_buffer.resize_destructive(aligned_width * surface_height);
|
||||||
chroma_buffer.resize(aligned_width * surface_height / 2);
|
chroma_buffer.resize_destructive(aligned_width * surface_height / 2);
|
||||||
|
|
||||||
// Populate luma buffer
|
// Populate luma buffer
|
||||||
const u8* luma_src = frame->data[0];
|
const u8* luma_src = frame->data[0];
|
||||||
|
|
|
@ -4,8 +4,9 @@
|
||||||
#pragma once
|
#pragma once
|
||||||
|
|
||||||
#include <memory>
|
#include <memory>
|
||||||
#include <vector>
|
|
||||||
#include "common/common_types.h"
|
#include "common/common_types.h"
|
||||||
|
#include "common/scratch_buffer.h"
|
||||||
|
|
||||||
struct SwsContext;
|
struct SwsContext;
|
||||||
|
|
||||||
|
@ -49,8 +50,8 @@ private:
|
||||||
/// size does not change during a stream
|
/// size does not change during a stream
|
||||||
using AVMallocPtr = std::unique_ptr<u8, decltype(&av_free)>;
|
using AVMallocPtr = std::unique_ptr<u8, decltype(&av_free)>;
|
||||||
AVMallocPtr converted_frame_buffer;
|
AVMallocPtr converted_frame_buffer;
|
||||||
std::vector<u8> luma_buffer;
|
Common::ScratchBuffer<u8> luma_buffer;
|
||||||
std::vector<u8> chroma_buffer;
|
Common::ScratchBuffer<u8> chroma_buffer;
|
||||||
|
|
||||||
GPUVAddr config_struct_address{};
|
GPUVAddr config_struct_address{};
|
||||||
GPUVAddr output_surface_luma_address{};
|
GPUVAddr output_surface_luma_address{};
|
||||||
|
|
Loading…
Reference in a new issue